Image Classification on ImageNet 1k (val) (Top-5 Accuracy)
90.664Top-5 Accuracymemorization
Evaluation results
| Method | Links | |
|---|---|---|
| memorizationFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.664 | |
| influence maxFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.59 | |
| EL2N (1 model)Fraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.534 | |
| supervised prototypesFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.528 | |
| self-supervised prototypesFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.466 | |
| active learningFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.354 | |
| EL2N (20 models)Fraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.352 | |
| forgettingFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.352 | |
| EL2N (20 models)Fraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.348 | |
| active learningFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.314 | |
| EL2N (1 model)Fraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.31 | |
| self-supervised prototypesFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.245 | |
| influence sum-absFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.224 | |
| memorizationFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.209 | |
| forgettingFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.12 | |
| supervised prototypesFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 90.076 | |
| randomFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.882 | |
| memorizationFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.837 | |
| EL2N (1 model)Fraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.818 | |
| randomFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.816 | |
| influence maxFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.804 | |
| EL2N (20 models)Fraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.698 | |
| influence sum-absFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.558 | |
| forgettingFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.486 | |
| self-supervised prototypesFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.46 | |
| DDDFraction of data kept=0.9, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.428 | |
| active learningFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.36 | |
| supervised prototypesFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.358 | |
| active learningFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.334 | |
| forgettingFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.112 | |
| randomFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.108 | |
| DDDFraction of data kept=0.8, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.06 | |
| memorizationFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 89.032 | |
| supervised prototypesFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.925 | |
| influence maxFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.684 | |
| DDDFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.682 | |
| EL2N (20 models)Fraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.674 | |
| influence sum-absFraction of data kept=0.7, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.618 | |
| self-supervised prototypesFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.547 | |
| DDDFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.484 | |
| EL2N (1 model)Fraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.476 | |
| randomFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.206 | |
| influence sum-absFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.082 | |
| influence maxFraction of data kept=0.6, Backbone=ResNet-50, Training framework=VISSL2022.06 | 88.062 |